AI-era builds

Client ventures under the Dago Digital contract needed working software — not slide decks — delivered without an in-house engineering team to build it. The work was mobile apps, a commerce platform, and retrieval pipelines, delivered end-to-end across AWS, Vercel, and Cloudflare Workers, whatever each engagement demanded rather than one chosen stack.

One build was native iOS and Android apps for a medtech venture, designed and built end-to-end. The venture ultimately could not raise its funding round. It ended.

Another build was a custom e-commerce platform, the whole way down: storefront, backend, payment processing, and the analytics layer needed to see whether any of it was working. No off-the-shelf commerce suite — every layer was a decision made for that specific client rather than configuration of someone else’s defaults.

Another build was graphRAG pipelines on Neo4j — retrieval-augmented generation grounded in an explicit graph of entities and relationships rather than flat vector similarity alone, for clients whose data had real structure worth preserving. The judgment call was recognizing when that structure was worth the extra modeling effort and when it was not.

Spread across three cloud providers by client-by-client necessity rather than choice, the infrastructure work was harder to reuse than it should have been. Given the choice again, standardizing on one platform early would have paid for itself many engagements over.